Several thoracic Injury Criteria (IC) and Injury Risk Functions (IRFs) have been proposed to relate THOR-50M chest deflections to the risk of rib fracture based on post mortem human subject tests. This study assessed the IC R_max, PCScore, D_max, DcTHOR, TIC_NFR and TIC_NSFR and their associated IRFs for their ability to predict the rib fracture risk of human occupants in stochastic population-based frontal impact accident reconstructions based on data from NASS/CDS. A THOR-50M finite element model was positioned in an occupant compartment model parameterised with respect to its interior geometry and restraint configuration. In addition, crash pulse characteristics and restraint activation times were parameterised, and 1,000 stochastic simulations were run. For the evaluated IC, the IRF values from each simulation were aggregated to a continuous curve with respect to delta velocity (ΔV). In general, the evaluated ICs and their IRFs overpredicted the risk of injury compared with rib fracture risk curves from the real-world crash data, except TIC_NSFR which was found to be too insensitive. The criterion that best matched the real-world crash data was PCScore, which had the closest match to the real-world data for the velocity at which 0.5 probability of AIS3+ rib fracture risk was predicted.